Home // International Journal On Advances in Software, volume 11, numbers 3 and 4, 2018 // View article
Authors:
Taketoshi Inaba
Chihiro Shibata
Kimihiko Ando
Keywords: CSCL; coding scheme; deep learning methods, automatic coding
Abstract:
In computer-supported collaborative learning research, it may be a significantly important task to figure out guidelines for carrying out an appropriate scaffolding by extracting indicators for distinguishing groups with poor progress in collaborative process upon analyzing the mechanism of interactive activation. And for this collaborative process analysis, labelling for appropriately representing properties of each contribution (coding) and statistical analysis are often adopted as a method. But as far as this paper is concerned, it tries to automate this huge laborious coding work with deep learning technology. In its previous research, supervised data was prepared for deep learning based on a coding scheme consisting of 16 labels according to speech acts. In this paper, with a multi-dimensional coding scheme with five dimensions newly designed aiming at analyzing collaborative learning process more comprehensively and multilaterally, an automatic coding is performed by deep learning methods and its accuracy is verified. The results indicate with certainty that we can introduce this model to authentic educational settings and that even for large classes with many students, we can perform real-time monitoring of learning process or ex-post analysis of big educational data. However, presenting raw results of automatic coding on each dimension is not enough to indicate the collaborative process quality to teachers and students. Therefore, a new rating model that can assess and visualize the quality of collaborative process is proposed.
Pages: 335 to 346
Copyright: Copyright (c) to authors, 2018. Used with permission.
Publication date: December 30, 2018
Published in: journal
ISSN: 1942-2628